Date: Fri, 18 Aug 2017 22:27:56 +0000
From: Vallish Vaidyeshwara <vallish@amazon.com>
To: Richard Cochran <richardcochran@gmail.com>
CC: <davem@davemloft.net>, <shuah@kernel.org>, <netdev@vger.kernel.org>,
        <linux-kernel@vger.kernel.org>, <eduval@amazon.com>,
        <anchalag@amazon.com>, <tglx@linutronix.de>
Subject: Re: [PATCH RESEND 0/2] enable hires timer to timeout datagram socket
Message-ID: <20170818222756.GB28737@amazon.com>
References: <1503081850-10671-1-git-send-email-vallish@amazon.com>
 <20170818201854.xes246oviptinwvq@localhost>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Disposition: inline
In-Reply-To: <20170818201854.xes246oviptinwvq@localhost>
User-Agent: Mutt/1.5.20 (2009-12-10)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1129
Lines: 34

On Fri, Aug 18, 2017 at 10:18:54PM +0200, Richard Cochran wrote:
> On Fri, Aug 18, 2017 at 06:44:08PM +0000, Vallish Vaidyeshwara wrote:
> > There has been a behavior change in 4.9 kernel with refactoring of Kernel
> > timer wheel in 4.8. We have a use case wherein our datagram socket
> > application is sensitive to socket timeout including long timeouts.
> > 
> > One of the test runs with a timeout value of 180 seconds timed out at
> > 190 seconds.
> 
> So the whole premise of the new timer wheel is that long timeouts need
> not be very accurate.
> 
> > Patch 1: Has core code change of enabling hires timer to timeout datagram
> > 	 socket on AF_UNIX and AF_INET domain
>

Hello Richard,

> Using hrtimers will hurt performance for most applications.  Can you
> please explain your use case and why is it so important?
>

We have a on-demand application that uses long timeouts and needs to react to
events within milliseconds. With this change in behavior, we notice from 4.4 to
4.9, the legacy application is now exhibiting incorrect behavior with same set
of system calls.

Thanks.
-Vallish

> Thanks,
> Richard
> 
>